New User Special Price Expires in

Let's log you in.

Sign in with Facebook


Don't have a StudySoup account? Create one here!


Create a StudySoup account

Be part of our community, it's free to join!

Sign up with Facebook


Create your account
By creating an account you agree to StudySoup's terms and conditions and privacy policy

Already have a StudySoup account? Login here


by: Brendan Kling


Marketplace > Texas A&M University > Geography > GEOG 661 > DIG IMAGE PROC ANALY
Brendan Kling
Texas A&M
GPA 3.78

H. Liu

Almost Ready


These notes were just uploaded, and will be ready to view shortly.

Purchase these notes here, or revisit this page.

Either way, we'll remind you when they're ready :)

Preview These Notes for FREE

Get a free preview of these Notes, just enter your email below.

Unlock Preview
Unlock Preview

Preview these materials now for free

Why put in your email? Get access to more of this material and other relevant free materials for your school

View Preview

About this Document

H. Liu
Class Notes
25 ?




Popular in Course

Popular in Geography

This 8 page Class Notes was uploaded by Brendan Kling on Wednesday October 21, 2015. The Class Notes belongs to GEOG 661 at Texas A&M University taught by H. Liu in Fall. Since its upload, it has received 21 views. For similar materials see /class/225766/geog-661-texas-a-m-university in Geography at Texas A&M University.

Similar to GEOG 661 at Texas A&M

Popular in Geography




Report this Material


What is Karma?


Karma is the currency of StudySoup.

You can buy or earn more Karma at anytime and redeem it for class notes, study guides, flashcards, and more!

Date Created: 10/21/15
Hyperspectral Remote Sensing Mo Bands Background What is Hyperspectral Remote Sensing Hyperspectral simply refers a large number of spectral bands usually in the Visible to midinfrared wavelengths 04 to 3 pm The instruments that acquire hyperspectral data are typically referred to as imaging spectrometers Hyperspectral Instruments AVIRIS Airborne Visible Infrared Imaging Spectrometer MODIS Moderate Resolution Imaging Spectroradiometer CASI 2 Compact Airborne Spectrographic ImagerZ EOl Daedelus Instruments MAS MODIS Airborne Simulator MASTER MODISASTER A Hyperspectral Data Cube Hyperspectral Profiles 5 w m o 500 moo o 500 woo Fig132 Line profile display created from hyperspectral data a139mnsect through portion of a hyperspeclral image cyscnle display ofspecmzl band horizontally versus position in the image Vertically cColoumd versiono Data Redundancy Are 256 bands really 32X better than 8 bands 0 Unfortunately probably not as there is much redundant information both spectrally and spatially we will consider the spectral redundancy 0 One way to do this is to examine a covariance or correlation matrix Correlation Matrices Figi 1341 a The correlation matrix for 196 wavebands39 covering 400 nm to 2400 nm for the A m L l anon of 0 b The remix of edge detecting the cmrclalion matrix From ENVI for 50 band Cuprite Data Set gt A Hyperspectral Image of Cu 0 rit evada EFFORT Corrected G 7 2 2008 mm B 7 2 3402 um We can also look at Principal Components Band lEigenvalue of variant 11 235076 86 89 29 2 18930 239 719 3l 5491 1363 2 09 41 1485 0514 0 56 5l 480 62984 0 18 61 395 30189 015 71 224 30048 0 09 8l 159 96694 0 06 91 117 00425 0 04 101 91150776 0 03 First Principal Component 7 quotl 7 i 397 Minimum Noise Fraction MNF Often used to determine the inherent dimensionality of the data to segregate noise in the data and to reduce the computational requirements for subsequent processing Green et al 1988 As implemented in ENVI essentially a cascaded PC transformations Minimum Noise Fraction II Step I PC transformation based on an estim aged noise covariance matrix decorrelates and rescales the noise in the data This results in transformed data that have unit variance and no band toband correlations Step II A standard PC transformation of the noisewhitened data The inherent dimensionality can be determined by examination of the final eigenvalues and associated images One set of images has large eigenvalues and coherent eighenimages and the second complementary set has nearunity eigenvalues and noisedominated images Problems with Dimensionality Remember as the dimensionality of a data set increases the number of training pixels per spectral class needed to preserve the class statistics increases At some point adding bands is not helpful unless more training pixels per class are availbable separating surface separating surface se arahrvg surface generated frorn the generated fmm he enersted from ME wallamlngplxels dssirad faurtrainingpixes desired many rainmgpixals desired a sulface n surface a u 0 El 5 39 l I u a o w o 39 C o 0 o E S 0 D D D o o 0 U5 0 o o I u 6 a u I training data 0 D testing data a c Fig 135 I i 1 honed training pixels generate a surface that also performs well for tesling da1ac Phenomena Hughes Hughes Phenomenon Classification Accuracy 8 If I 10 20 so 40 Number of Features Fig 116 T1 x 39 39 wnn increasing data dunensionalily This graph is the result of a four category classification an sure Atmospheric Correction Need for Atmospheric Correction Assume the image being imaged by a hyperspectral system has a uniform 100 spectral response and there is no atmosphere In this case What Will be imaged is the solar spectrum including Fraunhofer absorption lines if spectral sampling is ne enough Need for Atmospheric Correction If an atmosphere is added the spectrum recorded is modified by atmospheric absorption features Thus if we are interested in surface features it is necessary to remove tha atmsopheric absorption features as well as the shape of solar spectrum and atmospheric scattering effects sun deleclar via Fm l srspecrum acted 9 eat salsrspsctmm a 100 re ecting surface sun detector 1 spectrum detected snowmg arrears of mar spectrum atmosphere I 100 re ecting surface sun detector a DE reluctance spam m af surface msrwtsd by solar 9037 quantum and stmusnhsric arrears C real surface Fig133 Torma lou of the re ect snlar spectral il39mdiance atmosphcx of a given surface and Ike biasing arrow of the ion and scattering Atmospheric Correction Measures Compensation for the shape of the solar spectrum The measured radiances are divided by solar irradiances above the atmosphere to obtain the apparent re ectance of the surface Compensation for atmospheric gaseous transmission and molecular and aerosol scattering enable apparent re ectances to be converted to scaled surface re ectances Scaled surface re ectances can be converted to real surface re ectances after accounting for topographic effects or by assuming that surfaces are lambertian Flat Field Correction Used to normalize images to an area of known uniform at re ectance The radiance spectrum from this area is assumed to be composed primarily of atmospheric effects and the solar spectrum The reference spectra is divided into the spectrum of each pixel in the image resulting in an apparent re ectance that can be compared to laboratory spectra IARR 0 An average spectrum for the entire scene is computed and used as a reference spectrum Apparent re ectance is calculated by dividing the average spectrum into the spectrum of each pixel Empirical Line Calibration Image data is forced to match to selected field spectra Roberts et al 1985 Two or more ground targets are identified and re ectance is measured in the field Ideally targets consist of a light and a dark area Targets are identified in the image and spectra are extracted from an ROI Linear Regression is calculated between the field spectra and the image radiance to determine the transform from radiance to re ectance in each band Applied on a pixel by pixel basis ATREM Atmosphere Removal Program Gao et al 1992 Built upon SS Removes atmospheric water absorption in AVIRIS on a pixel by pixel basis using the 094 pm and 114 mm water vapor features Uses an approach called 3channel ratioing RADIANCE SEALED REFLECTANCE 3 I P n i 15 13 20 22 2 WAVELENGTH gm 0 04 05 08 10 112 14 Fig137 mm A 39rpnM hm V from Gao e aL 1993 mm permission from Elsevier Science permission Empirical Flat Field Optimized Re ectance Transformation EFFORT A relatively automated implementation of the Flat Field Calibration Method B oardman and Huntington 1996 AVIRIS spectra that match a loworder polynomial using a leastsquares fit are chosen to represent featureless spectra These spectra are averaged and a small gain is applied to remove systematic coherent noise present in every spectrum including small residual atmospheric effects near 20 pm caused by C02 Applied to data corrected using ATREM to remove residual effects Continuum Removal ENVIPlutWinduw v 4 File Edit Upwions 39 C 39 M wwm v 39 r39 Kl 1 39v 1 x V wl I 3 NM VJ 397 g L Continuum 3 Removed CC 05 Spectrum In I 39139 g g 5 015 Cun tinuum Vt mx x x ch 2 1 3 x 115 I rr39 x 139J Z 3 39 IlauJinite Spectrum a Ll W 1 It I a 1 ll chl o nql h life vcrrbcl 1239s Figure 144 Example of a med conrr39n uum and a continuumremoved spectrum for the mineral kaollnire Example Radiance Spectra Sp 1 ctral Library Viewer File Edit Options PlutFunctiIm quot11 I Li h m 1quot v I E Field 1 n EmilIII IZI Atmospheric Correction Comparison File Edit Options PlotFunction L E L39 E 4 Processing Binary Encoding F5 hltngt0 E if X01 lt Threshold then 7 7 hn 0 else hn 1 l x Threshold is typically averaged brightness m value for pixel m g spectrum Wavelength nm Fig 139 l39ummlmn om smlplc binan code for mAvuus gtpx c11u111 i More Advanced 10031 l6115 22199 Wavelength nm m m a E 500 E a m g 50 39 u F V g NBS 3 11112 7 1 a U gt c c 395 am 7 m i wow l 3947 10031 Wavelength nm Fig 13 1 o of 11m specuum Spectral Feature Fitting Uses absorption features for matching image spectra to reference library spectra Developed by Clark and other at USGS Methods vary but all require that a continuum be removed from spectra prior to analysis A particular absorption feature is isolated and a continuum mathematical function is used to isolate the feature Often the highpoints surrounding the absorption feature of interest are identi ed and a straightline fit between points is used to normalize the feature by dividing the continuum into original spectra Spectral Feature Fitting Continued A least squares fit is then calculated on a band by band basis and the total RMS error is used to provide an indication of goodness of fit Also possible to give some indication of depth of feature by determining what multiplicative factor is required to make the reference spectrum match unknown spectrum 7 High values 7 deep feature 7 Low values 7 shallow feature Spectral Analyst Finds closest match between a spectra of interest and library spectra THE ENVI implementation uses several techniques Binary Encoding Spectral Angle Mapper and Spectral Feature Fitting to come up with t of library spectra to reference spectra DO NOT USE BLlNDLY Example Spectral Fitting mm mm um swam Analyst IE File npuans Example lt Selected Spectra of Interest Regions of Interest ROIs Aimee Vsmu VZaam vm Spectral Feature Fitting Image RMS image for Alunite Bright 7 good fit Dark 7 Bad Fit Spectral Angular Mapper SAM In N dimensional space a pixel vector has both a magnitude and a direction angle measured with respect to the axes that de ne the coordinate space 0 The SAM technique attempts to identify a pixel only using the angular information SAM Illustration class 1 class 1 39 band 2 angles gt 7 class 2 angles lt 3 o pixel represented angle x Y Fig18 a Representing pixels by their angles from the band axes b Segmentmg he mum spectral space by angle ENVI implementation Determines the similarity of two substances by calculating the spectral angle between them Insensitive to illumination Generates a Rule Image 7 actual distance in radians between sample and each endmember Thresholds can be applied for classi cation SAM Example 739 H White 7 Zeolites Green 7 Calcite 7 Yellow Alunite 39 ed Kalonite ark Green IlliteM0nt Blue 7 Silica I aroon Buddingtonite Spectral Mixing An Andean Example


Buy Material

Are you sure you want to buy this material for

25 Karma

Buy Material

BOOM! Enjoy Your Free Notes!

We've added these Notes to your profile, click here to view them now.


You're already Subscribed!

Looks like you've already subscribed to StudySoup, you won't need to purchase another subscription to get this material. To access this material simply click 'View Full Document'

Why people love StudySoup

Bentley McCaw University of Florida

"I was shooting for a perfect 4.0 GPA this semester. Having StudySoup as a study aid was critical to helping me achieve my goal...and I nailed it!"

Amaris Trozzo George Washington University

"I made $350 in just two days after posting my first study guide."

Steve Martinelli UC Los Angeles

"There's no way I would have passed my Organic Chemistry class this semester without the notes and study guides I got from StudySoup."

Parker Thompson 500 Startups

"It's a great way for students to improve their educational experience and it seemed like a product that everybody wants, so all the people participating are winning."

Become an Elite Notetaker and start selling your notes online!

Refund Policy


All subscriptions to StudySoup are paid in full at the time of subscribing. To change your credit card information or to cancel your subscription, go to "Edit Settings". All credit card information will be available there. If you should decide to cancel your subscription, it will continue to be valid until the next payment period, as all payments for the current period were made in advance. For special circumstances, please email


StudySoup has more than 1 million course-specific study resources to help students study smarter. If you’re having trouble finding what you’re looking for, our customer support team can help you find what you need! Feel free to contact them here:

Recurring Subscriptions: If you have canceled your recurring subscription on the day of renewal and have not downloaded any documents, you may request a refund by submitting an email to

Satisfaction Guarantee: If you’re not satisfied with your subscription, you can contact us for further help. Contact must be made within 3 business days of your subscription purchase and your refund request will be subject for review.

Please Note: Refunds can never be provided more than 30 days after the initial purchase date regardless of your activity on the site.